管理层语调视角下的上市公司财务风险预警——基于FA-SVM模型的研究

Financial Risk Warning from the Perspective of Management Tone——Research Based on FA-SVM Model

为了解决现有财务风险预警模型中指标大量冗余相关性高、实时分析能力差以及忽略财务文本信息的问题,提出一种管理层语调视角下的财务风险预警模型。选取2016—2020年276家上市公司作为研究样本,从五个维度筛选了财务数据指标后,对企业财务年报中的前瞻性信息进行中文分词处理并根据常用情感词典计算其语调,并加入财务指标,利用因子分析降维,消除指标冗余和其间的相关性,提取出少量共性因子,最后采用支持向量机构建风险预警模型。结果表明,五维指标体系与因子分析所得的共性因子具有较强的一致性,证明了因子分析的经济意义可解释性。同时,结合财务文本信息可以有效提高财务风险预测的准确率,验证了财务文本对风险预警模型的有效性。

In order to solve the problems of the existing financial risk early warning models with a large number of indicators,high redundancy,poor real-time analysis capabilities,and ignoring financial text information for corporate bankruptcy prediction,this paper proposes a financial risk early warning model under the perspective of management tone.Selecting 276 listed companies from 2016 to 2020 as the research sample,after preliminary screening of financial data indicators from five dimensions,the forward-looking information in the corporate financial annual report is processed in Chinese word segmentation and the tone is calculated according to the commonly used emotional dictionary,and the financial data is added for indicators.And factor analysis is used to reduce dimensionality,to eliminate indicator redundancy and the correlation between them.We try to extract a small amount of common factors,and finally use support vector machines to build a risk early warning model.The experimental results show that the five-dimensional index system selected in this paper has strong consistency with the common factors obtained by factor analysis,which proves the economic significance of factor analysis and the interpretability.At the same time,the combination of financial text information can effectively improve the accuracy of financial risk prediction,which verifies the effectiveness of financial text on risk early warning models.